Bioinformatics researchers at the University of California, Santa Cruz, have conducted an audacious project to recreate the genome sequence of an ancestral mammal that lived 70 to 80 million years ago.

If it all sounds a little like the plot of Michael Crichton's bestseller Jurassic Park, that's not surprising. However, lead investigator David Haussler stresses his motives are driven by interest in evolution rather than economics. He is no John Hammond, and this is no Cretaceous-Tertiary Park.

Since the sequencing of the human genome, there has been intense interest in the field of comparative genomics - aligning corresponding stretches of DNA sequence from different species to monitor rates of substitution and other genetic changes during the course of evolution. A prime example is work of Eric Green (National Human Genome Research Institute), who collaborated with Haussler's team in the new study. Green has analysed 1.1 million bases of the same stretch of chromosomal DNA from 19 mammals, flanking the cystic fibrosis gene.

Based on the information in that dataset, Haussler and former postdoc Mathieu Blanchette, now at McGill University, decided to see if they could recreate this part of the genome of the common ancestor of all placental mammals, going back some 80 million years. The creature was a small shrew-like animal found in Asia. They succeeded, claiming 98 per cent accuracy. The work is published in the December issue of the journal Genome Research.

The project "sounds implausible," Haussler admits, "but there's enough information to reconstruct the ancestral genome on the basis of mammals that live today. We just need to sequence the genomes of these living mammals."

In Jurassic Park, Crichton speculated that scientists could recreate dinosaur DNA based on prehistoric blood samples extracted from mosquitoes that had been trapped in tree sap, and retrieved from amber. Because DNA is somewhat unstable, the recovered gene fragments were stitched together with linker sequences from modern-day frog DNA, to build a complete functional genome.

"What we come up with is not going to take us anywhere near Jurassic Park," insists Blanchette. "There are many reasons why the ancestral genome could not be put in a living organism -- technical reasons, it is difficult to synthesize long DNA sequences. But the main thing is we show we can reconstruct the ancestor with 98 per cent accuracy, and that's nice -- but its not enough to make that organisms livable. There will be a few [DNA] errors in crucial places."

Blanchette used an approach dubbed 'computerised paleogenomics' essentially to conduct a feasibility study, attempting to recreate in silico the DNA sequence of the 'boreoeutherian' mammalian common ancestor. This sequence differs from humans by about 30 per cent.

Blanchette wrote a computer program called TBA that aligned the 19 mammalian CF-region genome sequences, including human, chimpanzee, gorilla, mouse, dog, and cat, charting positions where DNA bases had been substituted, inserted or deleted, and enabling the group to determine which species had a common ancestor. "Then we try to figure out the evolutionary events that could produce the events that led to the alignment," says Blanchette.

Simulation studies show that the reconstructed ancestral sequence could account for the evolution of present-day mammalian species with an estimated 95 to 96 per cent accuracy, based on the available sequence. With additional mammalian sequences at hand, including bat and elephant, Blanchette says that accuracy would reach 98 per cent. The authors write that "in coming years, such reconstructions may help in identifying and understanding the genetic features common to eutherian mammals and may shed light on evolution of human or primate-specific traits." Some very intriguing examples, such as a point mutation in the FOXP2 gene that may have been relevant in the development of human speech, have already been documented.

Blanchette says he is now preparing to extrapolate this approach to the entire genome. "There's no big technical issue in scaling up to the whole genome," he says. "There are some regions that are not well behaved, that evolve much faster, that we can't reconstruct. But a large fraction will be quite accurate."